Sediment source tracing in a lowland agricultural catchment in southern England using a modified procedure combining statistical analysis and numerical modelling

• Published in: The Science of the total environment Vol. 414; pp. 301 - 317
• Main Authors: Collins, A.L., Zhang, Y., McChesney, D., Walling, D.E., Haley, S.M., Smith, P.
• Format: Journal Article
• Language: English
• Published: Netherlands Elsevier B.V 01.01.2012
• Subjects: Agriculture; Agronomy; Catchments; Discriminant Analysis; England; Farm tracks; Farming; Farms; Genetic algorithms; Geologic Sediments - chemistry; Geological Phenomena; Models, Chemical; Optimization; Road verges; Sediment sourcing; Sediments; Statistics, Nonparametric; Uncertainty; Urban; British Isles, England; ANE, British Isles, England; Urban; Farm tracks; Uncertainty; Agriculture; Road verges; Sediment sourcing
• ISSN: 0048-9697; 1879-1026
• DOI: 10.1016/j.scitotenv.2011.10.062

Catchment erosion, soil losses and resulting sediment pressures continue to represent cause for concern with respect to the ecological vitality and amenity value of riverine systems, including those in the agricultural catchments of southern England. Given that the sources of fine-grained sediment are typically diffuse in nature, it is essential to adopt a catchment-wide perspective to corresponding management strategies and sediment source tracing procedures have proved useful in assisting such planning. There remains, however, scope for further refining sediment sourcing procedures and on that basis, a recent study in the upper River Kennet (~214km2) catchment in southern England, provided an opportunity for designing and testing a refined statistical procedure for sediment source discrimination with composite fingerprints using Genetic Algorithm (GA)-driven Discriminant Function Analysis, the Kruskal–Wallis H-test and Principal Components Analysis. The revised statistical verification of composite signatures was combined with numerical mass balance modelling using recent refinements including a range of tracer weightings and both local and GA optimisation. Comparison of the local and global optimisation increased confidence in the outputs of local optimisation and the goodness-of-fit for the predicted source contributions using the optimum composite signatures selected from the revised statistical testing ranged from 0.914 to 0.965. Overall relative frequency-weighted average median source type contributions were estimated to be 4% (agricultural topsoils; predicted deviate median inputs 1–19%), 55% (unmetalled farm track surfaces; predicted deviate median inputs 9–91%), 6% (damaged road verges; predicted deviate median inputs 4–42%), 31% (channel banks/subsurface sources; predicted deviate median inputs 5–41%) and 4% (urban street dust; predicted deviate median inputs 0–20%). The study provides further evidence of the importance of eroding farm tacks as a catchment scale sediment source and confirms the utility of tracing for assembling information on sediment inputs from both the agricultural and urban sectors. ► We describe the application of a modified composite sediment source fingerprinting procedure. ► The procedure includes a new approach for verifying composite fingerprints. ► Agricultural and diffuse urban sources are included in the source apportionment. ► Findings reveal the catchment scale importance of farm track and channel bank/subsurface sources. ► Uncertainty is explicitly recognised.